AI Token Price
NVIDIA

2 current models

NVIDIA API pricing

NVIDIA API pricing: Nemotron 3.5 Lightning $0.025 / $0.12 and Nemotron 3 Super $0.085 / $0.40 per 1M input / output tokens. Checked 13 Oct 2026 against NVIDIA.

Current NVIDIA models

Current NVIDIA models and prices per million tokens
Batch API model ID
$0.085 $0.40 — 256K 11 Mar 2026 — nemotron-3-super-120b-a12b
$0.025 $0.12 — 1M 11 Aug 2026 — nemotron-3.5-lightning

What a month costs on NVIDIA models

Monthly cost of three example workloads on current NVIDIA models, at list prices
ModelChat app10M input, 2M output tokensSupport bot50M input (70% cached), 5M outputCoding agent750M input (85% cached), 8M output
Nemotron 3 Super$1.65$6.25$66.95
Nemotron 3.5 Lightning$0.49$1.85$19.71

Cheapest third-party host prices checked 13 Oct 2026, general-purpose models, prompts under any long-prompt threshold, no batch discount. Cached input is billed at each model's cache-read price; cache-write charges aren't included. Price your own traffic · Agent cost per task

How NVIDIA prices its API

  • Nemotron models are open weights; prices shown are from third-party hosts.

NVIDIA model IDs

API model IDs, context windows and release dates for current NVIDIA models
ModelAPI model IDContextMax outputReleased
Nemotron 3.5 Lightningnemotron-3.5-lightning1M—11 Aug 2026
Nemotron 3 Supernemotron-3-super-120b-a12b256K—11 Mar 2026

Questions

How much does the NVIDIA API cost?

NVIDIA API pricing: Nemotron 3.5 Lightning $0.025 / $0.12, Nemotron 3 Super $0.085 / $0.40 per 1M input / output tokens, checked 13 Oct 2026. Every model is billed per token, separately for input and output.

Does NVIDIA charge less for cached input?

NVIDIA does not list a cached-input price for its current models on its pricing page.

Is there a batch discount?

NVIDIA does not list a batch discount on its pricing page as of 13 Oct 2026.

NVIDIA model families

Pricing rules

  • Nemotron models are open weights; prices shown are from third-party hosts.